Purchase Predicting with Click Stream Data

Abstract

Everyday using ecommerce sites many customers purchase different types of product form online. This virtual shops are now biggest business now. One of the biggest challenge of the ecommerce sites is showing ads and offers to the correct customer. If they can do this than their sell rate will increase. For this reason customers clickstream data is rich source of customer behavior analysis. The aim of this project is to develop such a functional classifier which can predict purchaseevents correctlyasmoreaspossible. In this project, I use multiple linear regression technique to predict purchase event. In this technique I build a scoring model with multiple linear regression. Then I set some thresholds manually and use it with different size of dataset. Then I pick the best acting threshold compacting precision, recall, accuracy.I also use ROC graph to verify the threshold.